Download When ‘More’ in Statistical Learning Means ‘Less’ in Language:

When ‘More’ in Statistical Learning Means ‘Less’ in Language:
Individual Differences in Predictive Processing of Adjacent Dependencies
Jennifer B. Misyak ([email protected])
Morten H. Christiansen ([email protected])
Department of Psychology, Cornell University, Ithaca, NY 14853 USA
Abstract
Although statistical learning (SL) is widely assumed to play a
key role in language, few empirical studies aim to directly and
systematically link variation across SL and language. In this
study, we build on prior work linking differences in
nonadjacent SL to on-line language, by examining individualdifferences in adjacent SL. Experiment 1 documents the
trajectory of adjacency learning and establishes an individualdifferences index for statistical bigram learning. Experiment 2
probes for within-subjects associations between adjacent SL
and on-line sentence processing in three different contexts
(involving embedded subject-object relative-clauses, thematic
fit constraints in reduced relative-clause ambiguities, and
subject-verb agreement). The findings support the notion that
proficient adjacency skills can lead to an over-attunement
towards computing local statistics to the detriment of more
efficient processing patterns for nonlocal language
dependencies. Finally, the results are discussed in terms of
questions regarding the proper relationship between adjacent
and nonadjacent SL mechanisms.
Keywords: Predictive Dependencies; Sentence Processing;
Bigrams; Serial Reaction Time; Artificial Grammar
Introduction
With the expansion of studies on statistical learning (SL)
over the past decades, focus has intensified towards probing
the potential role for probabilistic sequence learning
capabilities in acquiring and using linguistic structure (e.g.,
Gómez, 2002; Saffran, 2001). A clearer understanding has
in turn begun to crystallize about the ways in which SL
mechanisms may underpin language across various levels of
organization—phonetic, lexical, semantic, syntactic—and
across differing timescales—phylogenetic, ontogenetic, and
microsecond unfoldings. Largely missing from this picture,
however, is empirical evidence that directly links language
and SL abilities within the typical population.
There are, though, a few recent studies that address the
issue of whether better statistical learners are indeed better
processors of language. In a small-scale study of individual
differences, Misyak and Christiansen (2007) observed that
standard measures of SL performance are positively
associated with comprehension accuracy for various
sentence-types
in
natural
language.
Conway,
Bauernschmidt, Huang and Pisoni (2010) reported that
better SL performance correlates with better processing of
perceptually-degraded speech in highly-predictive lexical
contexts. Misyak, Christiansen and Tomblin (2010) found
that more-skilled statistical learners of nonadjacent structure
were also more adept at the on-line processing of longdistance dependencies in natural language. Thus far, these
results would support the general assumption that SL and
language processes are systematically interrelated, with
positive correspondence in intraindividual variation across
them. But is it always the case that greater SL is associated
with better language functioning? Or, may excelling at one
of these implicate poorer performance at the other?
Such ability-linked reversals in performances within a
cognitive domain would not be unprecedented. As an
example, bilingual individuals appear to possess more
efficient ‘inhibitory control’ processes than their
monolingual peers across a number of studies, which has
usually been imputed in some manner to bilinguals’ greater
experience with ‘control’ processes for suppressing
irrelevant information in the course of successfully using
two languages (see Bialystok et al., 2004). However, in a
negative priming paradigm where distractor locations that
were supposed to be previously ignored became relevant for
facilitating responses to a current trial (as they do for
monolinguals), bilinguals are at a disadvantage in the
cognitive control task, with decreases from a neutral
baseline in performance accuracy (Trecanni et al., 2009).
Analogously then, might there be natural language contexts
in which superior SL skill also becomes disadvantageous?
One possibility is that a statistical learner may focus too
much on computing certain statistics, while ignoring others,
with repercussions for their linguistic processing. For
example, language embodies predictive dependencies that
can be broadly characterized as involving either adjacent or
nonadjacent temporal relationships. Thus, a good adjacency
learner might perform poorly on nonadjacent dependencies
in language. Introducing a new task for documenting microlevel trajectories and individual differences in SL, Misyak et
al. (2010) were able to link variation in nonadjacent SL
positively to signature differences in reading time patterns
for the complex nonlocal dependency structure of centerembedded object-relative clause sentences. However, this
study raises a new set of questions, including ones that
directly bear on the above hypothetical, namely: Does the
timecourse of adjacent SL differ from that of nonadjacent
SL? Can substantial differences in adjacent SL also be
empirically related to on-line sentence processing? And if
so, might this differ from the kinds of positive correlations
observed for nonadjacency processing?
We investigated these questions by adapting the AGLSRT paradigm from Misyak et al. (2010) to isolate the
learning of adjacent dependencies. The task implements an
artificial grammar (AG) within a modified two-choice serial
reaction-time (SRT) layout, using auditory-visual sequence-
2686
strings as input. Experiment 1 thus documents the group
trajectory and range of individual differences for adjacency
learning obtained from this task. A ‘bigram index’ reflecting
individual differences in adjacency learning is then used to
probe relationships to the processing patterns observed in
our subsequent natural language experiment (Experiment 2).
Experiment 1: Statistical Learning of
Adjacencies in the AGL-SRT Paradigm
The ability of humans to use adjacent statistical information
has been demonstrated across various studies. As early as
two months of age, humans can identify bigrams, or firstorder adjacent pairs, from the co-occurrence frequencies of
elements within a constrained temporal sequence (Kirkham,
Slemmer & Johnson, 2002). Throughout later development
and adulthood, humans can also use adjacent conditional
probabilities to locate relevant constituent-boundaries in a
continuous stream composed of nonwords, tones, visual
elements, or nonlinguistic sounds (see Gebhart, Newport &
Aslin, 2009, for a review). And further, both children and
adults can learn adjacent predictive dependencies that signal
the underlying phrase structure of an artificial language
(Saffran, 2001).
Below, we adapt the biconditional grammar of Jamieson
and Mewhort (2005) to examine adults’ SL of bigrams. This
grammar was chosen since it is defined by first-order
transitions only, imposes no positional constraints on
element placement, and generates strings of equal length.
These merits thereby permit us to effectively isolate the
learning of predictive adjacencies by our participants.
Method
Participants Thirty native English speakers from the
Cornell undergraduate population (15 females; age: M=19.4,
SD=0.8) were recruited for course credit.
Materials Participants observed sequences of auditoryvisual strings generated by an eight-element grammar in
which every element could be followed by one of only two
other elements, with equal probability. Each string consisted
of 4 elements, with adjacent probabilities between them as
shown in Table 1.The nonwords (jux, tam, hep, sig, nib, cav,
biff, and lum) were randomly assigned to the stimulus
tokens (a, b, c, d , e, f, g, h) for each participant to avoid
Table 1: Transition probabilities for elements at positions n
and n + 1 of a string, with n as an integer from (0, 4).
Element at position n +1 of string
Element
at n
a
b
c
d
e
f
g
h
a
b
c
d
e
f
g
h
0
0
0
0
0
0
.5
.5
.5
0
0
0
0
0
0
.5
.5
.5
0
0
0
0
0
0
0
.5
.5
0
0
0
0
0
0
0
.5
.5
0
0
0
0
0
0
0
.5
.5
0
0
0
0
0
0
0
.5
.5
0
0
0
0
0
0
0
.5
.5
0
Figure 1: The pattern of mouse clicks for a single trial
with the auditory target string “jux cav lum nib.”
potential learning biases due to specific sound properties of
words. Auditory versions of the nonwords were recorded
from a female native English speaker and length-edited to
550 ms. Written versions of nonwords were presented with
standard spelling in Arial font (all caps) and appeared within
the rectangles of a 2 x 4 computer grid (see Figure 1). Each
of the 4 columns of the computer grid, from left to right,
displayed the nonword options corresponding to the 1st thru
4th respective elements of a string. Ungrammatical strings
were created by introducing an incorrect element at the 2nd
or 3rd string position, with the next element being one that
legally followed the incorrect one (e.g., as in “a *d e g”).
Procedure Each trial corresponded to a different
configuration of the grid, with each of the eight written
nonwords centered in one of the rectangles. Every column
contained a nonword (target) from a stimulus string, as well
as a foil. The first column contained the selection for the
first element of a string, the second column contained the
selection for the second element, and so on. For example, a
trial with the stimulus string jux cav lum nib, as shown in
Figure 1, might contain the target jux and the foil hep in the
first column; the target cav and the foil biff in the second
column; the target lum and the foil sig in the third column;
and the target nib and the foil tam in the fourth column.
Each nonword appeared equally often as target and as foil
within and across the columns. The top/bottom locations of
targets and foils were randomized and counterbalanced.
Participants were informed that the purpose of the grid
was to display their selections and that a computer program
randomly determines a target’s location within either the top
or bottom rectangle. On every trial, participants heard an
auditory stimulus string composed of four nonwords and
were instructed to respond to each nonword in the sequence
as soon and as accurately as possible by using the computer
mouse to select the rectangles displaying the correct targets.
Thus for any given trial, after 250 ms of familiarization to
the visually presented nonwords, the first nonword of a
string (the target) was played over headphones. Next, the
second, third, and fourth words of a given string were each
played after a participant had responded in turn to the prior
nonword. For example, on a trial with the stimulus string jux
cav lum nib, the participant should first click the rectangle
containing JUX upon hearing jux (Fig. 1, left), CAV upon next
hearing cav (Fig. 1, center-left), LUM upon hearing lum (Fig.
2687
1, center-right), and NIB upon hearing nib (Fig. 1, right).
After a participant had responded to the last nonword, the
screen cleared for 750 ms before a new trial began.
An intended consequence of this design is that, for any
given trial, the first element of a string cannot be anticipated
in advance of hearing the auditory target. However, all
subsequent string transitions might be reliably anticipated
using statistical knowledge of the bigram structure. Thus, as
participants become sensitive to the bigrams, they should be
able to anticipate the string transitions, which should be
evidenced by faster response times (following standard SRT
rationale). Accordingly, our dependent measure on each trial
was the reaction time (RT) for a predictive target, subtracted
from the RT for the non-predictive initial-column target
(which serves as a baseline and controls for practice
effects). The predictive target used in this calculation was
equally distributed across all non-initial columns across
trials. Analogously, for an ungrammatical string trial, if
participants are sensitive to the bigrams, then their RTs for
incorrect, or violated, elements should be slower; thus, the
DV for ungrammatical trials was the RT for the illegal
target subtracted from the initial-target RT.
There are 64 unique strings (8 x 2 x 2 x 2) defined by the
grammar; these were all randomly presented once each for
each grammatical block of trials. Training consisted of six
grammatical blocks, followed by an ungrammatical block of
16 trials and then a single grammatical (‘recovery’) block.
Transitions across blocks were seamless and unannounced.
After these eight blocks, participants were informed that
the strings had been generated according to rules specifying
the ordering of nonwords and were asked to complete two
tasks involving prediction and bigram recognition,
respectively. The prediction task consisted of 16 trials that
were procedurally similar to the trials observed during
training, but with the omission of the auditory target for the
final column.1 Instead, participants were told to select that
nonword in the final column that they believed best
completed the sequence.
In the bigram task, participants were randomly presented
with 32 test items of auditory nonword-pairs. They were
requested to judge whether each pair followed the rules of
the grammar by pressing ‘yes’/’no’ computer keys. Half of
the test items were the 16 bigrams licensed by the grammar
(e.g., a b); the remaining half were illegal pairings formed
by reversing each bigram (e.g., b a). Thus, successful
discrimination reflects knowledge of the conditional
bigrams, rather than only sensitivity to co-occurrences.
Results and Discussion
Analyses were performed on only ‘good’ trials—that is,
accurate string-trials with only one selection for each target.
1
Instructing participants to complete string endings allows for
maximal procedural similarity to the speeded training trials without
introducing additional cue prompts that would be needed if the
aurally-omitted element varied across non-initial columns. It also
avoids any indirect feedback effects from presenting the next
element after a participant’s correct/incorrect medial selection.
Figure 2: Group learning trajectory (mean RT difference
scores per block) and accuracy for prediction (left bar) and
bigram (right bar) tasks.
Prior to analysis, the data from five participants were
omitted (2 for withdrawing participation; 2 for improperly
performing the task, with less than 40% good trials; and 1
for abnormally elevated RTs, averaging in excess of 1470
ms per single response). For remaining participants, good
trials averaged 88.2% (SD=5.9) of training block trials.
Mean RT difference scores, as described above (i.e., for
grammatical trials: initial-target minus predictive-target RT;
for ungrammatical trials: initial-target minus illegal-target
RT) were computed for each block and submitted to a oneway repeated-measures analysis of variance (ANOVA) with
block as the within-subjects factor. Since the assumption of
sphericity was violated (χ2(27) = 113.27, p <.001), degrees
of freedom were corrected using Greenhouse-Geisser
estimates (ε = .33). Results indicated a main effect of block
on RT difference scores, F (2.31, 55.36) = 3.82, p =.02. As
seen in Figure 2, mean RT difference scores appear to
increase by the final training block, decrease in the
ungrammatical block, and increase once again in the
recovery block. As RT difference scores measure the
amount of facilitation from the predictive targets, an
improvement in scores across blocks (as seen here) reflects
sensitivity to the adjacent dependencies.
Planned contrasts between the ungrammatical block and
preceding/succeeding grammatical blocks confirmed a
performance decline for the ungrammatical trials (Block 6
minus Block 7: M= -42.0 ms, SE=19.6, t(24) = 2.14, p =.04;
Block 8 minus Block 7: M= 39.8 ms, SE=17.8 ms, t(24) =
2.23, p =.04). This provides evidence for participants’
learning of the sequential dependencies, consistent with
standard interpretations in the sequence learning literature
for comparing RTs to structured versus unstructured
material (e.g., Thomas and Nelson, 2001).
Since the amount of exposure to the dependencies during
training is equivalent to that which a similar number of
participants (n=30) received in the Misyak et al. (2010)
study of nonadjacent SL, this invites a comparison of group
learning trajectories. The RT timecourse pattern
documented here for adjacent SL is very similar to that
observed for nonadjacent SL, but with greater variance in
2688
performance for the final training block and with ostensibly
more modest (albeit not statistically different) performance
in the recovery block. In both cases, sensitivity to the
statistical structure does not show signs of emerging until
after considerable exposure (the 5th block of training).
Mean accuracy on the prediction task was 55.3%
(SD=17), which was not above chance (t(24) = 1.51, p
=.14)—despite 20% of participants scoring at or above 75%.
However, accuracy on the bigram task reflected adjacency
learning (t(24) = 4.66, p <.0001), with a mean of 57.6%.
This performance level is consistent with participants’
judgment accuracy in an AGL study with manipulations of
this same type of grammar when participants are tested with
ungrammatical items containing few rule violations
(Jamieson & Mewhort, 2009). Bigram scores further ranged
from 37.5 – 71.9%, but with less variance (SD=8) than that
observed in the prediction task. In post-study questioning,
only four participants disclosed that they had noticed any
general pattern in the sequence but were unable to verbalize
at least one instance of a bigram, suggesting that their
performance in the bigram task was not the product of
explicit recall or well-formulated meta-knowledge. Next, we
use scores on this bigram index to assess whether and how
variation in adjacent SL may be associated with differences
in processing local and nonlocal language dependencies.
Experiment 2: Individual Differences in
Language Processing and Statistical Learning
Sensitivity to both local and long-distance relationships is
indispensable to processing natural language, and pervades
basic aspects of our everyday sentence comprehension and
production—such as those involved in relating the modified
subject/object of a described action or state to the main
event of a sentence (embedded relative clauses), in
identifying whether someone is the recipient or doer of an
action (agent-patient thematic roles), and in correctly
linking subjects with their verbs (number agreement). The
aim of Experiment 2 is to investigate whether predictive
processing as exemplified by adjacent SL is empirically
related to the on-line processing of such natural language
contexts. Consider the following examples of the sentencetypes that constitute the focus of the current experiment.
(1a-b) The reporter [that attacked the senator / that the
senator attacked] admitted the error.
(2a-b) The [crook/cop] arrested by the detective was
guilty of taking bribes.
(3a-b) The key to the [cabinet/cabinets] was rusty from
many years of disuse.
In the first sentence example, the subject-relative (SR; 1a)
and the object-relative (OR; 1b) versions differ with respect
to the manner in which the embedded verb attacked relates
to its object. This involves a more complex, backwardstracking long-distance dependency (to the head-noun) for
ORs. In prior studies using materials resembling those in
(1a-b), greater processing difficulty is elicited at the main
verb of ORs compared to that of SRs, with considerable
individual differences in the magnitude of this effect (e.g.,
Wells, Christiansen, Race, Acheson & MacDonald, 2009).
Next, consider the sentence pair (2a-b), which is
temporarily ambiguous between a main verb (MV) and a
reduced relative (RR) clause interpretation. Its resolution is
influenced by the constraint of thematic fit—the fit between
the head noun phrase (the crook or the cop) and the verbspecific roles of the verb (arrested). Given verb-specific
conceptual knowledge, the reader knows that cop is a
typical agent of arrested, whereas crook is a typical patient.
Controlling for animacy, thematic fit functions as an
immediately integrated constraint computed over the noun
and adjacent verb—with its effect on RTs occurring in the
subsequent agent NP region (McRae, Spivey-Knowlton &
Tanenhaus, 1998). Thus, the second condition (2b) in which
the initial noun is a typical agent for the adjacent verb will
elicit greater processing difficulty for the RR interpretation
than that for the corresponding patient condition (2a). For
our purposes, this provides an example of sensitivity to a
local relation relevant for on-line sentence processing.
Lastly, (3a-b) illustrate subject-verb number agreement.
In English, it is required that a number-marked subject (key)
agrees with the number-marking of its verb (was). This is
the case irrespective of the numerical marking of any
intervening material (e.g., to the cabinet/s), and individuals
are sensitive to this fact during reading. When a sentence’s
head noun is singular, individuals read longer at the MV in a
condition where the ‘distracting’ local noun (cabinets)
mismatches in number (i.e., is plural) than in a condition
where the local noun matches the head noun’s number (i.e.,
is singular); shorter reading latencies are also found for the
word after the verb in the match condition (Pearlmutter,
Garnsey & Bock, 1999). Although subject-verb agreement
may occur locally between adjacent constituents, materials
in the literature (and here) have involved a nonlocal
dependency created from interposing a prepositional phrase.
Method
Participants The same participants from Exp. 1 participated
directly afterwards in this experiment for additional credit.
Because the analyses reported below involve correlations
with the bigram index from Exp. 1, data was omitted for
those participants already excluded in Exp. 1 analyses and
from three others (2 for bilingual status and 1 for declining
to participate in the second task).
Materials There were four sentence lists, each consisting of
9 practice items, 60 experimental items, and 50 filler items.
The experimental items were sentences drawn from
previous studies of sentence processing: 20 subject-object
relative clauses (SOR; Wells et al., 2009), 20 reduced
relative ambiguities influenced by thematic fit (TF; McRae
et al., 1998), and 20 subject-verb agreement transitives (SV; Pearlmutter et al., 1999). A yes/no comprehension probe
followed each item. Item conditions within sentence sets
were counterbalanced across lists.
Procedure Each participant was randomly assigned to a list,
whose items were presented in random order using a
2689
a standard word-by-word, moving window, self-paced
reading paradigm. Millisecond reading times (RTs) per
word and accuracy were recorded for analyses.
Results and Discussion
Overall comprehension accuracy across participants was
high, M= 87.4%, SD=7.6. RTs in excess of 2500 ms (0.2%
of data) were removed, and remaining RTs were then
length-adjusted for the number of characters in a word using
a standard procedure (Ferreira & Clifton, 1986). Unless
otherwise noted then, all RTs reported below for each of the
sentence sets have been length-adjusted, with the same
sentence regions examined as those in the original studies.
RTs connected with relevant effects for each of the sets
were then used to probe for associations with individuals’
bigram scores from Experiment 1, as summarized below.
Subject-Object Relatives. Results replicated the main
effect for clause-type at the MV from Wells et al. (2009),
F(1, 21) = 5.55, p= .03. OR MVs were read reliably longer
(91 ms) than SR MVs. However, there was no signification
correlation between bigram scores and MV RTs for either
SR (r = .04, p= .85) or OR (r = -.16, p= .47) sentences.
Thus, differences in adjacent SL did not appear to directly
map onto differences in processing long-distance
dependencies in these relative clauses.
Thematic Fit. The influence of TF was replicated at the 2word MV region (e.g., was guilty), F(1, 21) = 6.42, p =.02,
albeit not at the directly preceding agent NP region.2 Agent
conditions were read 39 ms longer than patient conditions at
the MV region. The correlation between bigram scores and
unadjusted RTs at the MV of the ‘congruent’ patient
condition was not significant (r = .29, p= .19); but for the
’incongruent’ agent condition, the correlation reached
marginal significance (r = .40, p= .06), with better adjacent
statistical learners taking longer to read the disambiguating
verb phrase. This suggests a tendency for greater bigram
sensitivity (in adjacent SL) to negatively correspond with
resolving nonlocal ambiguity when the local TF constraint
provides an opposing bias to the RR clause interpretation.
Subject-Verb Agreement. A 34 ms effect of match (i.e.,
the difference between match and mismatch conditions) was
obtained at the verb, F(1, 21) = 31.28, p< .0001, which
replicated Pearlmutter et al.’s (1999) findings. There was a
smaller effect of match (23 ms) at the post-verb region, F(1,
21) = 4.48, p= .05, which was also numerically present but
not reliable in Pearlmutter et al. Additionally, the correlation
between bigram scores and RTs was significant for the
effect at the verb (r = .51, p= .02), with better bigram
learning corresponding to a larger effect of match condition.
To further examine differences in processing patterns
according to SL status, a median-split was performed on
bigram scores, establishing 57.8% as the cut-off for defining
membership in either a “high” bigram (n=11, M= 63.9%,
2
The later-occurring but nonetheless reliable effect of thematic
fit is likely due to differences in the length of the moving window
used in this study (1-word) and that by McRae et al. (2-word).
Figure 3: RT patterns on the S-V agreement sentences by
bigram group (high/low) and condition (match/mismatch).
SD=4.0) or “low” bigram group (n=11, M= 51.4%, SD=5.8).
Significant bigram-group differences emerged for the effect
of match condition across regions (as shown in Figure 3).
While the low-bigram group did not elicit a significant
effect of match condition at either the verb or post-verb
region (p= .13 and p= .91, respectively), the high-bigram
group showed a clear effect in both regions (both p’s< .001).
As apparent in Fig. 3, the high-bigram group demonstrated
greater sensitivity to the interference created by the locally
mismatched marking of the noun in the prepositional phrase
(which was irrelevant for computing agreement). Thus, the
better adjacent SL of the high-bigram group was related to
generally less efficient processing than that by their lowbigram peers of the long-distance dependency entailed by
the initial noun and verb. Since bigram groups did not differ
in comprehension accuracy for any sentence-types in the
experimental sets (all p’s > .15), nor fillers (p= .83), these
RT patterns were not the result of a speed-accuracy tradeoff.
Our findings suggest that adjacent SL skill may not
directly tap into the processes most relevant for handling
long-distance dependencies in natural language—even
though nonadjacent SL abilities appear to do so. Thus, while
Misyak et al. (2010) reported a positive association between
differences in nonadjacent SL and processing for the same
SOR clauses as used here, no correlation was detected for
adjacent SL. More generally, this is consistent with the lack
of within-subjects correlation found between adjacent and
nonadjacent SL in Misyak and Christiansen (2007).
However, while ‘high’ bigram learners may not differ
from ‘low’ learners on processing long-distance relations as
such, their increased sensitivity to local relations might
interfere with the processing of the longer-distance elements
within the sentence. This tendency is seen in the TF set,
where above-average bigram tracking abilities seem to have
a negative effect for processing the MV—the site where the
initial, nonlocal ambiguity must be resolved. Similarly, too
much sensitivity to local information is clearly evidenced
within the last sentence set, where the irrelevant marking of
an adjacent noun negatively affects better bigram learners’
resolutions of S-V agreement, with protracted RTs also at
the MV site of integrating the long-distance dependency.
2690
General Discussion
References
This study investigated the processing of adjacent predictive
dependencies to address questions related to the timecourse
of adjacent SL and the nature of any empirical association to
natural language variation. While a learning trajectory
similar to nonadjacent SL was documented in Exp. 1,
findings from Exp. 2 indicated that above-average gains in
adjacent SL performance do not necessarily translate to
gains in language processing. Notably, those individuals
who were strongly attuned to tracking statistical bigrams
exhibited a negative pattern of correlations to tracking
longer-distance aspects of language when either
countervailing adjacent constraints or nearby distractive
elements were present. This inverse pattern was not
evidenced, though, when processing long-distance relations
without conflicting local information (in the SOR clauses).
Instances where better bigram learners were worse
language processors (or tended towards less efficient RT
patterns) occurred when the integration of adjacent
information (between a head-noun and part-participle verb)
induced greater difficulty for resolving an ambiguity as a
RR (the TF constraint in Exp. 2)—or when locally irrelevant
information disrupted agreement computations between a
nonlocal subject and verb (S-V agreement in Exp. 2). It
would appear in these situations that those better in adjacent
SL, although excelling at bigram pattern recognition in the
SL task, are overly attuned to adjacency patterns and
become more susceptible to local ‘garden-paths’; in such
cases, it may be the ‘over-focus,’ rather than any preexisting
weakness in processing long-distance dependencies (as
evidenced by parallel performance of groups in the SOR set)
that hinders efficient resolution of nonlocal relationships.
This interpretation of our findings suggests that
intraindividual differences in processing biases for the
integration of competing constraints among adjacent- and
nonadjacent dependencies may contribute to variation
across SL-linked language processing skills. As such, it
speaks to an open issue regarding whether different systems
or different processing biases may be entailed by adjacent
and nonadjacent processing capabilities in humans. It has
been proposed, for instance, that the two forms of
processing may be subserved by separate brain areas
(Friederici et al., 2006), or that the two types of SL are only
nominally distinct as the outcome of task-specific attention
processes that may selectively hone in on adjacent or
nonadjacent statistics (cf. Pacton & Perruchet, 2008). The
findings here, of negative and specific associations between
adjacent SL and aspects of language processing, suggest that
future individual differences research incorporating careful
attention to a diversity of natural dependency-structures may
be needed to help establish the proper relation between these
two manifestations of SL and the extent to which they may
‘tap’ into the same underlying mechanisms.
Bialystok, E., Craik, F.I.M., Klein, R. & Viswanathan, M. (2004).
Bilingualism, aging, and cognitive control: Evidence from the
Simon task. Psychology and Aging, 19, 290-303.
Conway, C.M., Bauernschmidt, A., Huang, S.S. & Pisoni, D.B.
(2010). Implicit statistical learning in language processing:
Word predictability is the key. Cognition, 114, 356-371.
Ferreira, F. & Clifton, C. (1986). The independence of syntactic
processing. Journal of Memory and Language, 25, 348-368.
Friederici, A.D., Bahlmann, J., Heim, S., Schibotz, R.I. &
Anwander, A. (2006). The brain differentiates human and nonhuman grammars: Functional localization and structural
connectivity. Proceedings of the National Academy of Sciences,
103, 2458-2463.
Gebhart, A.L., Newport, E.L. & Aslin, R.N. (2009). Statistical
learning of adjacent and nonadjacent dependencies among
nonlinguistic sounds. Psychonomic Bulletin & Review, 16, 486490.
Gómez, R. (2002). Variability and detection of invariant structure.
Psychological Science, 13, 431-436.
Jamieson, R.K. & Mewhort, D.J.K. (2005). The influence of
grammatical, local, and organizational redundancy on implicit
learning: An analysis using information theory. Journal of
Experimental Psychology: Learning, Memory, and Cognition,
31, 9-23.
Jamieson, R.K. & Mewhort, D.J.K. (2009). Applying an exemplar
model to the artificial-grammar task: Inferring grammaticality
from similarity. Quarterly Journal of Experimental Psychology,
62, 550-575.
Kirkham, N.Z., Slemmer, J.A. & Johnson, S.P. (2002). Visual
statistical learning in infancy: Evidence for a domain general
learning mechanism. Cognition, 83, B35-B42.
McRae, K., Spivey-Knowlton, M.J. & Tanenhaus, M.K. (1998).
Modeling the influence of thematic fit (and other constraints) in
on-line sentence comprehension. Journal of Memory and
Language, 38, 283-312.
Misyak, J.B. & Christiansen, M.H. (2007). Extending statistical
learning farther and further: Long-distance dependencies, and
individual differences in statistical learning and language. In
Proceedings of the 29th Annual Cognitive Science Society (pp.
1307-1312). Austin, TX: Cognitive Science Society.
Misyak, J.B., Christiansen, M.H. & Tomblin, J.B. (2010).
Sequential expectations: The role of prediction-based learning in
language. Topics in Cognitive Science, 2, 138-153.
Pacton, S. & Perruchet, P. (2008). An attention-based associative
account of adjacent and nonadjacent dependency learning.
Journal of Experimental Psychology: Learning, Memory, and
Cognition, 34, 80-96.
Pearlmutter, N.J., Garnsey, S.M. & Bock, K. (1999). Agreement
processes in sentence comprehension. Journal of Memory and
Language, 41, 427-456.
Saffran, J.R. (2001). The use of predictive dependencies in
language learning. Jrnl of Memory and Language, 44, 493-515.
Thomas, K.M. & Nelson, C.A. (2001). Serial reaction time
learning in preschool- and school-age children. Journal of
Experimental Child Psychology, 79, 364-387.
Treccani, B., Argyri, E., Sorace, A. & Della Sala, S. (2009).
Spatial negative priming in bilingualism. Psychonomic Bulletin
& Review, 16, 320-327.
Wells, J.B., Christiansen, M.H., Race, D.S., Acheson, D.J. &
MacDonald, M.C. (2009). Experience and sentence processing:
Statistical learning and relative clause comprehension.
Cognitive Psychology, 58, 250-271.
Acknowledgments
Thanks to Parry Cadwallader, Becky Fortgang and Stephan
Spilkowitz for assistance with running participants.
2691